Deferring performance conversations does real organizational damage. Most managers know this and do it anyway, because the conversations are uncomfortable and the short-term cost of avoidance is invisible. A team that carries persistent underperformance quietly accumulates drag – on morale, on output, on the expectations of everyone watching. Acting on performance issues, and doing it with clarity and honesty, is one of the harder parts of the job that actually matters. On that, there’s no real argument.

The question is what signals you use, and what you do with them.

Code quality and PR output are visible, measurable, and genuinely informative. They’re a reasonable starting point. But they’re a starting point, not the whole picture. A lot of the most valuable engineering work doesn’t show up in a dashboard. The senior engineer who spent a week reviewing an architectural decision and preventing a costly mistake. The person who wrote documentation that onboarded three new hires without requiring anyone’s time. The engineer who built the relationship with the product team that meant technical concerns actually got heard. None of that is captured by PR frequency.

There’s also a risk in treating output metrics as objective when they’re not. PR size and frequency depend on what someone is working on. A complex infrastructure change might be one large PR that took three weeks of careful work. A developer on a mature codebase will look different in the metrics from one building on unstable foundations. Context matters, and metrics without context are routinely misread.

The advice to make difficult firing decisions quickly is partly right. Clarity is better than ambiguity, for everyone involved. Kim Scott’s work in Radical Candor makes the case well – the manager who avoids hard feedback out of misplaced kindness isn’t being kind. They’re allowing someone to fail slowly and publicly instead of helping them course-correct. That’s true.

But “act quickly” without the full sentence – act quickly once you have clarity, have given honest feedback, and have given a reasonable opportunity to improve – creates its own problems. Managers who act too quickly on incomplete signals get it wrong. They lose people they shouldn’t lose. They create a culture where people are afraid to struggle visibly, which means problems go underground rather than getting solved. And in most employment contexts, moving fast without a documented process creates legal exposure that falls squarely on the company.

The pragmatic approach

The goal of talent assessment is a high-performing team, and how you build that matters as much as the destination.

Use multiple signals. Output metrics are one input among several. How does someone respond to feedback? Do they make the people around them better? Are they growing, even if slowly? These are harder to quantify and easier to dismiss as soft, but they’re predictive in ways that PR counts aren’t.

Give feedback early and specifically. The most common failure mode in performance management isn’t acting too slowly – it’s not being honest enough soon enough. By the time someone is on a formal plan, most good managers wish they’d been clearer six months earlier. Clear, specific, early feedback gives people a genuine chance to respond and often resolves the situation without escalation.

Document as you go. Not to build a case, but because it forces clarity. If you can’t write down what the performance issue is, what a good outcome would look like, and what support you’ve offered, you probably don’t have enough clarity to act yet.

There’s no formula for this. Talent assessment done well is contextual, continuous, and honest. Metrics help, but they don’t replace judgment.